RealTime IT News

Brian Stevens, CTO, Red Hat

Brian StevensRed Hat has long been considered the leading enterprise Linux vendor. It's a position that Red Hat CTO and Vice President Engineering Brian Stevens is aiming to maintain.

Since its inception, Red Hat has striven to challenge the existing software ecosystem. First with its namesake Red Hat Linux, which was discontinued in 2004 (and now continued in its Fedora Core distribution), as well as with its Red Hat Enterprise Linux distribution which is currently at version 4.

The Linux vendor recently acquired middleware vendor JBoss in a $350 million deal that received mixed reviews from analysts.

Red Hat is now preparing its next major enterprise release, which is expected to be highlighted by its virtualization functionality.

Stevens recently sat down with internetnews.com to talk about what's coming from Red Hat and the challenges that it faces.

Q: What's happening with RHEL 5 development? Is Alpha 1 out now?

RHEL 5 is currently targeted for release by the end of year. We actually took a different model this time.

We're convinced that there is a better way to develop software, so what we did is we blew up the notion of an Alpha and we use Fedora as an alpha. The engineers are goaled on not just producing enterprise quality software, but driving it through upstream in terms of the community.

In the early days it was about providing a version of Linux that is differentiated somehow; instead it's now about how do we participate in the upstream projects through Fedora.

FC 5 and FC6 constituted the role of an alpha and now we're going right to Beta 1 in a few weeks. By and large the release driver is virtualization, which ripples everywhere and is much more than just a hypervisor.

Q: Beyond the Virtualization hypervisor, what are you focusing your virtualization efforts on in RHEL 5?

We're trying to provide a really good user experience, so the first thing we did is do define an API. Xen doesn't have an API it allows users, IT and even end users to build application to it.

So we started something called libvirt.org -- which is not just a Red Hat thing -- which took a cut at an API. Then we started building out on top of that.

We think virtualization is for more than just servers, it's pretty powerful for just an end user. On top of libvirt we're creating we're building all the GUI capability to quickly create a VM {virtual machine}, lifecycle management, resource migration, memory, CPU and monitoring.

Instead of just giving users the tools to manage virtualization, we're also building an API that allows other people to innovate.

What we're doing with para-virt ops is around a compatibility layer between a guest operating system and the hypervisor.

Right now there is no well-defined API between the Linux kernel and the hypervisor itself.

With VMware the API has always been you've got to work with a piece of x86 hardware. So our operating system running on VMware didn't have to develop to a different API.

We just make it work on Intel and if we get that right then it should work in theory and we just have to test and certify.

Xen is very different and that's why it's disruptive. With para-virtualizaiton Xen, let's tell the operating system that we're virtualizing it and if we let it know that it can help us.

Now we need to have a model on how we further innovate that API cause there are going to be new ideas. So now what's happing is VMware has absolutely seen the light of day and they want to support a para-virtualized version of their hypervisor.

Q: How is the JBOSS integration coming?

It's a hybrid integration model. There is a standalone division reporting to Matthew Szulik with Marc Fleury [JBoss founder] as a general manager.

We quickly realized from both sides that there are some things that are very natural to do. Everything we've built up in the last five years is about how we deliver enterprise software to customers, how to leverage our ISVs, our test grid matrix -- everything that happens.

We've spent millions of dollars on this process automation, and we can bring that to JBoss engineers. What we're finding is that integration is happening everywhere.

The engineering team at JBoss has the vision, has mature technology, and that part is just running. What we're wrapping around that is delivery; our sales team has already been integrated.

How can we as a service organization accelerate and add value to what they are already doing?